Chemical Engineering Journal 116 (2006) 85–95

Individual impeller flooding in aerated vessel stirred by

multiple-Rushton impellers
Andrej Bombač ∗ , Iztok Žun
Laboratory for Fluid Dynamics and Thermodynamics, Faculty of Mechanical Engineering,
University of Ljubljana, Aškerčeva 6, 1000 Ljubljana, Slovenia
Received 26 April 2005; received in revised form 18 October 2005; accepted 19 October 2005

Abstract
The paper presents flooding detection in the dispersion of air into water in a stirred vessel equipped with a multiple-turbine impeller. All
experiments were performed in a pilot-size mixing vessel using single, dual and triple Rushton turbine impellers. Deionized water and compressed
air were used as a working fluid. The flooding recognition method based on resistivity probe response was applied, defining impeller flooding as
the appearance of ragged cavities behind the blades of the individual impellers. Classifying the flooding regimes into a simple generalized flow
map gave the sequence of impeller flooding: spreading from the lowest impeller flooding subsequently to the upper impellers. A comparison of
our individual impeller-flooding data with the results found in the literature is shown for single impeller and partially for dual- and triple-turbine
impeller, where due to the lack of published data only a comparison for the lowest turbine was possible.
© 2005 Elsevier B.V. All rights reserved.

Keywords: Stirred vessel; Multiple impeller; Aerating; Impeller flooding

1. Introduction [11]. In the work of Alves and Vasconcelos [2] it is shown that
even in a flooding regime their mixing model and correlation of
Nowadays vessels with a high liquid aspect ratio are usually mixing time can be used for multiple impeller stirring. The find-
used in fermentation processes. Such fermenters are basically ings of Sardeing et al. [30] studying the aeration in the vicinity of
equipped with multiple impellers to improve hydrodynamic the liquid surface with a specially composed multistage impeller
characteristics and consequently mass and heat transfer. In broth showed a compromise choice between a high gas flooding rate
fermentation the rheological properties change from the begin- (partial flooding) and a long bubble plume length in a vessel.
ning to end of the process. A change from a Newtonian to a Among the first researchers studying the flooding phe-
pseudo-plastic fluid causes a change in viscosity of up to four nomenon, Mikulcova et al. [20] and Rushton and Bimbinet [29]
orders of magnitude, which strongly affects the bulk flow field described the term ‘flooding’. Since then several authors have
in a vessel. The question arises of how this transformation exerts researched this phenomenon. In the literature, the term flood-
an influence on flooding. This phenomenon is dependent on the ing has often been used to describe a transition to unsatisfactory
input rates of gas flow and power dissipated into the broth via operation of a stirrer in a tank, while the definition of flooding
two-phase impeller flow discharge. A low gas input flow rate is still not well established. In addition, many different exper-
causes insufficient dissolved oxygen in the liquid bulk, while an imental methods have been reported for the measurements of
excessively high gas flow rate may lead to the appearance of dead hydrodynamic regimes once flooding appears. As derived from
zones and flooding. If it still occurs, in some special occasions single impeller studies, these methods can be divided into two
it does not have any influence on, e.g. micromixing effects in an groups: (i) techniques based on global measurements (such as
aerated tank [15] nor does the aeration always affect the mixing gas holdup, power drawn) combined with visual observation of
number in the same way; it can either decrease or increase it two-phase flow circulation through the vessel wall and (ii) tech-
niques based on local two-phase characteristics such as fluid
velocity and local void fraction of the impeller discharge flow,
∗ Corresponding author. Tel.: +386 1 4771 406; fax: +386 1 4771 447. measured with a constant temperature anemometer, resistiv-
E-mail address: [email protected] (A. Bombač). ity probe, micro-propeller or vibrating vane. Furthermore, the

1385-8947/$ – see front matter © 2005 Elsevier B.V. All rights reserved.
doi:10.1016/j.cej.2005.10.009
86 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95

Nomenclature
b impeller blade length (m)
c impeller clearance above the base (m)
d impeller disk diameter (m)
D stirrer diameter (m)
f baffle width (m)
fb impeller blade frequency (s−1 )
Fl Flow number, q/(nD3 )
Fr Froude number, Dn2 /g
FrT tank Froude number, Tn2 /g
g gravity (m/s2 )
h distance between the impellers (m)
H liquid height (m)
ib number of impeller blades
k1 , k2 correlation coefficients
M torque (Nm)
Mo Morton number, gη4 /(σ 3 )
n impeller frequency (s−1 )
nb impeller blade frequency (s−1 )
P ungassed power (W)
q volumetric gas flow rate (m3 /s)
R regression coefficient
Fig. 1. Comparison of literature data regarding the term ‘flooding’.
s sparger clearance above the base (m)
T tank diameter (m)
w impeller blade width (m) no more horizontal flow of gas starting from the impeller and
We Weber number, ρn2 D3 /σ reaching the wall [29]. The flooding point was defined with a
z number of turbines sharp decrease of the holdup for a given total power, which
corresponds with visual observations of the flow patterns on
List of symbols the vessel wall. Mikulcova et al. [20] postulated the criterion
α local void fraction (%) for impeller flooding considering the hysteresis range, i.e. LFT
η viscosity (Pa s) and FLT in the dispersing of air into water with a single disk
ρ liquid density (m3 /kg) turbine. The method was based on the minimal impeller speed
σ surface tension (N/m) technique combined with conductivity measurements of the liq-
uid near the impeller. Wiedmann [35] and Nienow et al. [21]
detected the onset of flooding from changes of the gassed power
manner of achieving flooding should be described more pre- drawn; the flooding point was defined by step change in the
cisely according to the flow regime transition. Changing of gassed power curve under constant impeller speed and vary-
the flow regime from loading to flooding (LFT) or flooding ing gas rate. According to Warmoeskerken and Smith [34] the
to loading (FLT) corresponds to regimes which constitute hys- flooding transition reflects the unique states of the two-phase
teresis. Another question arises if the hysteresis is of negligible flow around the stirrer. Based on the qualitative determination
effect, as found by Warmoeskerken and Smith [33] or not, e.g. of the liquid radial outflow vector near the stirrer with a micro-
[1,14,17,20,22,27]. Such significant confusion found in the lit- propeller, a new definition of impeller flooding was introduced.
erature regarding the term flooding can be best seen in the flow In this regime there is an axial flush of gas through the impeller
regime map, see Fig. 1, which takes into consideration data for plane up to the free liquid surface with no radial discharge two-
a vessel of diameter 0.45 m and D/T equal to 0.33. It can be seen phase flow. Lu and Chen [16] studied the flooding phenomenon
that the FLT of devices at pilot or industrial scale at Fr = const is by using a conical hot-film anemometer. The onset of flooding
achieved at markedly higher Fl values than those at laboratory was postulated with sharp transition of r.m.s. turbulent fluctu-
scale regardless of whether the prediction includes the D/T ratio ation velocity of the impeller-discharge liquid flow. Later Lu
or not. and Ju [17] examined the gas-filled cavity formation, flooding
transition and pumping capacity of an aerated disk turbine. The
1.1. Single-turbine impeller flooding was determined on the basis of the impeller-discharge
fluid velocity change. According to Hudcova et al. [13,14] the
Using global measurements of gas holdup and power drawn effect of liquid height and impeller clearance on the FLT were
in air liquid mixing, flooding was defined as a state (varying studied based on changes of the gassed power curves. Only
the impeller speed at constant gas flow rate) when there was the change of liquid height itself had no effect on FLT, while
 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95 87

increasing the impeller clearance had a similar effect to the gas The purpose of this paper is to enable the simple estimation of
dispersion process with a ring sparger of diameter greater than individual impeller flooding for single-, dual- and triple-turbine
the impeller. Tatterson and Morrison [32], based on a retrofitting impellers in the dispersion of air into water. The experimen-
of Zweitering’s experimental data, analyzed the effect of tank tal method using resistivity probe response has been described
to impeller diameter ratio on LFT. Their model is an extension in detail in [6–8]. The method enabled the recognition of the
of the previous work of Warmoeskerken and Smith [33] and following cavity structures: vortex-clinging structure VC, the
Biesecker (as cited by Tatterson and Morrison [32]). Paglianti appearance of the first large cavity 1L, two large cavities 2L,
et al. [24] presented a simple method based on non-intrusive small  3-3 structure S33, large  3-3 structure L33 and ragged
detection of conductance fluctuations for the detection of FLT cavities RC, respectively. The RC structure corresponded to the
in a stirred tank. A later work [25] provided a mechanistic model flooding regime while the loading regime could form the L33 or
for predicting the LFT in an aerated vessel stirred by Rushton S33 structure.
turbine. The model neglects the influence of sparger size, as well In multiple-turbine stirring the lower turbine became flooded
as impeller clearance and liquid height. first, followed by the middle impeller and finally the upper
impeller successively by increasing the gas flow rate. The LFT of
1.2. Multiple-turbine impellers the lower turbine in the dual- and triple-turbine systems occurred
at noticeably lower gas flow rates than that in single-turbine stir-
A survey of the literature shows that only few studies have ring, while the upper impeller became flooded at higher gas flow
been published concerning flooding in multiple-turbine driven rates. Good agreement with flooding prediction data from the lit-
tanks. Henzler [12] presented the criterion based on dimension- erature was found for single-turbine stirring and for the lowest
less numbers of the LFT for single impeller stirring and for the impeller in dual- and triple-turbine stirring. Otherwise a compar-
lower impeller in multiple impeller stirring. The criterion was ison of our results with those from the literature was not possible
extrapolated from the measured data of other researchers on for the middle and upper impellers due to a lack of data in the
standard Rushton turbine stirring. In a study [31] the impeller literature.
gas-filled cavity structures were presented for a triple impeller.
Detection of cavities was performed with a vibrating vane 2. Experimental
mounted on the vessel wall; findings were obtained on the basis
of frequency analysis of pressure pulses of the vane. Structures A cylindrical flat-bottomed Perspex vessel of 450 mm ID
were classified into a flow regime map where a clear delineation with four baffles mounted perpendicularly to the vessel wall
between loading and flooding regimes was given for the low- was used. The geometrical details for the stirred vessel can
est impeller while for the middle and uppermost impellers it be found in Fig. 2. Deionized water and compressed air at
was only indicated. Nocentini et al. [23] reported qualitative room temperature were used as working fluids in all experi-
observations of two-phase flow field in a vessel with a four- ments. In multiple-turbine stirring the liquid height in the vessel
stage impeller. Only at a very low impeller speeds and at high was increased by the distance between turbines. Impeller speed
gas flow rates were all the turbines flooded; with an increase of was measured by IR-pulse transmitter with an absolute error
impeller speed the turbines became loaded sequentially. A sharp of ±1 rpm. An HBM transducer enabled torque measurements
transition of power drawn was found for the lowest impeller with error of ±0.02 Nm, while the gas flow rate was measured
whereas the transition was gradual and rather undefined for the with calibrated rotameters with the following errors: ±0.4 m3 /h
other turbines. In their study of air–water dispersion with dual for measuring tubes from 5 to 25 m3 /h, and ±0.14 m3 /h for mea-
impellers, Hudcova et al. [14] used a special shaft with two strain suring tubes from 1 to 8 m3 /h, respectively. At the highest flow
gauges, enabling individual impeller torque measurements. On rates two rotameters were connected in parallel. The experi-
this basis the flooding was detected for each impeller separately mental setup is depicted in Fig. 3 with the indicated optical
and was defined either from the step change in the gassed power window located on the vessel wall at the lower impeller height to
curve or visually. LFT appeared differently for each impeller enable visualization of the impeller discharge flow. A high-speed
and was strongly dependent on their clearance (h). Com- monochromatic video camera, a Weinberger VISARIO with a
pared to a single impeller the lower impeller required a higher Nikon 50/2 focus lens and a CMOS image sensor and optimized
speed to achieve a FLT (by h/D ≤ 1.5) otherwise when the full noise reduction was used in the case of dispersing with single
independent flow patterns are present the lower FLT can be esti- impeller. A high recording rate was needed at 4000 images/s at
mated reasonably well by the equation of Nienow et al. [21]. electronic shutter speed 200 ␮s. The image size format at such
Abrardi et al. [1] examined dissipated power and mixing time in a speed was set to 768 × 512 pixels, with a pixel size of 10 ␮m.
a sparged vessel driven with a dual turbine for different hydrody- Recording time span was up to 4 s which produced up to 12 Gb
namic regimes including the flooding of the lower impeller. The of data in a single experimental run
findings of Alves and Vasconcelos [2] are interesting, present- In order to obtain local information on gas filled cavity struc-
ing some new aspects of mixing in gas–liquid contactors with tures, a microresistivity probe of 11 ␮m tip size was used, located
triple impeller stirring. Some properties, e.g. mixing time, can in the close vicinity of the outer edge of the impeller blade. At
be still evaluated even in the flooding regime; otherwise more the lowest impeller speed at least 600 cavities were detected
data on flooding could not be obtained even in recent reviews: for further statistical treatment. Measurements were always per-
e.g. [10,18,26]. formed in the same manner starting from low to high impeller
88 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95

Fig. 2. Geometrical parameters of the vessel and stirrers.

speeds, with stepwise increasing of the gas flow rate at constant recognition have been described in detail in [6]. Local detection
impeller speed. The reproducibility and directional sensibility of the phases was performed by the resistivity probe producing
was assumed to be equal to those evaluated in [6]. voltage response as a corresponding structural function [38]:
The LFT can be detected by rapid reduction of void fraction

in a turbine discharge two-phase flow, as well by gas-filled cavity 1, x is occupied by p
structures change. Some experimental studies of gas-filled cavity Mp (x, t) = p = {L, G, S} (1)
0, x is occupied by p
structures and local void fraction distribution in water–air dis-
persion have been presented [6,7,8,16,17]. While Lu an Ju [17] In a two-phase flow field at a particular point x three phase states
studied gas-filled cavity configuration, flooding and pumping p are possible at any time t: the liquid phase L, gas phase G or
capacity using a constant temperature anemometer in single- phase interface S, respectively. Using a discrete fourier trans-
turbine stirring, in [7,8,37] gas-filled cavity structures and local formation of the structural function Mp enabled the presentation
void fraction based on resistivity probe response in an aerated of the significant frequencies of an appearing gas phase. The
vessel stirred with single and dual Rushton disk impellers were fourier coefficients Xk were obtained from:
investigated. Furthermore, in [5], attention was paid to gas-filled
cavity structures in aerating in a pseudo-plastic fluid such as N−1
 j2πik
a CMC-water dilution. Full details of the experimental work Xk = t Mp (tk ) e− N (2)
outlined here and the criterion for gas-filled cavity structure k=0
 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95 89

Fig. 3. Experimental setup.

where t denotes the time interval between successive instants was also quite evident in the L33 structure. From the evolution
ti . Among fourier coefficients Xk that correspond to the fre- of fourier coefficients in the frequency domain (with increasing
quency k/(Nt) only coefficients from k = 0 to (N/2) − 1 are gas flow rate) the increase of the coefficient at fb /2 is obvious.
meaningful. From the given criterion the following gas-filled At the point when both coefficients fb /2 and fb are nearly equal,
cavity structures were recognized: vortex-clinging structure the L33 structure was recognized, as depicted in Fig. 4.
(VC), structure with one large cavity (1L), structure with two The method used here was recently compared with two other
large cavities (2L), small  3-3 structure (S33), large  3-3 struc- experimental methods, i.e. the global gas holdup method and
ture (L33) and ragged cavity structure (RC). In Fig. 4 the resis- power drawn method in a standard mixing vessel equipped with
tivity probe responses and their fast fourier transformations are a single Rushton turbine [9] where the results were in good
shown for the above mentioned structures. The development of agreement among themselves. Given the fact that air-dispersing
the cavity structures follows from the top to bottom, i.e. from was performed with a single Rushton impeller, all the changes
the VC to RC structure, respectively. From the probe response at the local level (impeller) are reflected at the global one (gas
over longer periods of observation it was evident that large cav- holdup, power drawn) which actually enables the application of
ities were present on the same blades. The beginning of the S33 the global methods.
structure in the frequency-domain can be seen from the appear-
ance of a significant fourier coefficient at frequency fb /2, which 3. Results and discussion
increases in accordance with increasing of the gas flow rate. This
pattern was taken as a criterion of S33 structure recognition. The 3.1. Single-turbine impeller
L33 structure is a combination of three smaller and three larger
large cavities that is a combination of large cavities exclusively. Ragged cavities are in good agreement with the findings in
The alternating sequence of smaller and larger large cavities [21,31,33,34] where the delimitation between “loading” and
90 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95

Fig. 4. Cavity structures and corresponding probe response (a) and frequency transformation (b).
 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95 91

flooding, case (c). Here clouds of bubbles moved from the gas
sparger past the impeller upward to the free surface. Flooding
was recognized with the presence of the RC structure. Such a
hydrodynamic regime was marked (with the increasing of the gas
flow rate at constant impeller speed) with an adequate set of (nF
and qF ) as “impeller flooding” and the transition to flooding as
LFT, respectively. The corresponding correlation of the exper-
imental data was preferred in the same type as in [2,3,19,21]
to predict impeller flooding in single- and multiple- turbine
impellers:

FlF = k1 · FrFk2 (3)

where k1 , k2 and regression coefficient R are given in

Table 2. The experimental results are depicted in Fig. 6
as separate marks. At lower impeller speeds correspond-
ing to Fr < 0.08 the LFT occurred at the appearance of
RC structure deriving from VC structure while at higher
impeller speeds RC structure derived from  3-3 struc-
ture, which is in good agreement with previous results
[6].
To enable a comparison between experimental results and
the data from the literature, only correlations of LFT transition
are taken and depicted in Fig. 6. The geometrical configurations
with respect to D/T ratio, impeller spacing, sparger, etc. of the
experimental setups are shown in Table 1.
Regarding the vessel scale, as can be seen in Fig. 6, the fol-
lowing conclusions can be drawn: the predicted LFT appeared
in larger vessels than ours [12,35] at somewhat higher Fl values,
and in smaller vessels the predicted LFT appeared at lower Fl
values [17,20]. The experiment in [1] in which LFT appeared
at lower Fl values was also carried out in a smaller vessel than
ours.

Fig. 5. Photos of discharge two-phase flow in (a and b) loading and (c) flooding
regime, respectively.

“flooding” corresponding to  3-3 or large clinging cavity struc-

ture and RC was shown. Using a high speed camera in our
experimental work some distinctiveness in the form of discharge
two-phase flows can be seen from the side view as shown in
Fig. 5. Photos of the discharge two-phase flow were taken at
constant impeller frequency 4.45 s−1 and three different gas flow
rates; 6 m3 /h (a), 7 m3 /h (b) and 8 m3 /h (c), respectively. In case
(a) a radially oriented discharge flow was observed with typi-
cal L33 structures. Despite being under the same L33 structure,
increasing of the gas flow rate led to the change of the dis-
charge flow direction, i.e. from radial to upward, respectively,
as depicted in case (b). This state was found as the last possible Fig. 6. Comparison of our experimental LFT results in single-turbine stirring
loading regime, while a further increment of gas caused impeller with those from the literature.
92 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95

Table 1
Summary of the geometrical conditions in impeller flooding detection
Author Prediction Flow regime T (m) D/T c/T h/T Sparger Detection
 T −3.43
Mikulcova et al. FlF = 6.11±1.24 Fr
100
D Fr ≤ 0.9 0.19, 0.2 1/3 – Tube LFT, FLT
[20] 0.31 –0.39

Zlokarnik [36] FlF = 0.19FrF0.75 0.1 ≤ Fr ≤ 2 0.2, 0.2 1/3 – Tube LFT
0.45 –0.45

0.21F rF 2.1(D/T )
Henzler [12] FlF = + Re > 1 × 104 , Ind. 0.3 – – Ring, tube LFT
(T/D−2.04)1.3
0.14F rF 7.54(D/T ) Fr ≤ 5 scale –0.4
, (h/T ) > 0.75
(T/D−2.25)1.5
Wiedman [35] FlF = √  a 0.1 ≤ Fr ≤ 1.5 0.45 1/3 1/6 – Ring LFT
0.973 −0.15 + 4 Mo + 0.05 ln We
Fr Fr + –1.5
 −1
0.2Fr 1.6 , where a = 0.605 + ln We
Fr
Nienow et al. [21] FlF = 30(D/T )3.5 FrF FrF ≤ 0.85 0.29 0.33 0.25,0.4 – Ring FLT
–1.2 –0.5

Wamoeskerken FlF = 1.2FrF Fl ≤ 0.36 0.44 0.4 0.4 – Ring FLT

and Smith [34] –1.2

Hudcova et al. Exper. data Fl ≤ 0.16 0.56 1/3 1/3 0–1 Ring FLT
[14]
Smith et al. [31] Exper. data Fr ≤ 1, Fl ≤ 1 0.64 0.4 – 0.08–1 Two rings LFT
Tatterson and (FlF − 0.6FrF ) = −0.031FrT,F (T/D) + FLT
Mor-rison [32] 0.046, 2 ≤ FrT,F (T/D) ≤ 64, retrofitted
literature data
0.333
Lu and Ju [17] nF = A(T/D)1.167 (gq/D4 ) (4.07 + Fr ≤ 1.3, 0.28 0.25 1/3 – – Ring LFT, FLT
1.21ib − 0.147ib 2 ) A = –0.5
0.064(FLT) or 0.072(LFT)
Nocentini et al. Exper. data Fr ≤ 0.2 0.23, 0.2 1/3–3/5 – Ring LFT, FLT
[22] 0.39, –0.5
0.57
Nocentini et al. Exper. data Fr ≤ 0.2 0.23 0.52 0.5 1 Ring LFT, FLT
[23]
Abrardi et al. [1] Exper. data Fr ≤ 2.5, 0.39, 1/3 1/3 2/3 8 × 2 mm h LFT, FLT
Fl ≤ 0.2 0.65
Alves and FlF = 0.33FrF0.69 Fl < 0.25 0.292 1/3 1/2 1 Ring LFT
Vasconcelos [2]
Paglianti et al. Comp. code Fr ≤ 0.3, 0.2, 0.33 1/3 – Ring, p.plate FLT
[24] Fl ≤ 0.15 0.72

3.1.1. Multiple-turbine impeller bine occurs at higher gas flow rates than that of single-impeller
Defining the hydrodynamic regimes in multiple-turbine stir- stirring.
ring is more complex. Treating of the bulk liquid flow with one As the impellers separate, fully independent flow patterns
circulation loop per impeller is admissible only with h/D = 1.5 occur, and from the literature survey it seems that such cases
as proposed in [4,13,31], otherwise the impellers generate a were mainly studied and oriented toward the flooding of the
common interactive discharge flow. The LFT of the lower and lower turbine only. In such manner as proposed in [12], the LFT
upper turbines occurred similarly as in single-turbine stirring via of the lower turbine can be estimated reasonably well by using
different cavity structure development. In dual-turbine stirring, the same prediction as for single impeller stirring. The conclu-
the transition corresponded to the appearance of the RC struc- sions of Hudcova et al. [14] and Smith et al. [31] were similar,
ture developed from VC, 1L, 2L at Fr < 0.08, and from  3-3 both used the prediction of Nienow et al. [21] for single impeller
structures at higher Fr, respectively. But quite a big difference stirring. Thus the given estimations can be explained comparing
between the lower and upper turbine arises with respect to the the individual impeller flooding conditions of multiple-turbine
gas dispersing ability. As can be seen in Fig. 7, the upper turbine and single-turbine stirring as depicted in Fig. 7. The LFT of the
remained in a loading regime at markedly higher Fl values (when lower turbine (Du1) occurred at much lower gas flow rates than
it became flooded) than the lower one. This is in good agreement in single-turbine stirring. As can be seen from visual observation
with our previous findings in [8]. Otherwise in dual-turbine stir- through the vessel wall, two-phase circulation in the liquid bulk
ring with impeller spacing h/D = 1.5 a common interactive flow ceased due to reduced liquid pumping capacity of the lower tur-
was produced where the flooding transition of the lower tur- bine, so the liquid bulk in the close vicinity and especially below
 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95 93

Fig. 8. Experimental results: individual impeller flooding in triple-turbine stir-

Fig. 7. Experimental results: individual impeller flooding in dual-turbine stirring ring compared with those from the literature.
compared with those from the literature.
from VC to RC at Fr < 0.2 and from  3-3 to RC structure at
the lower turbine remained undispersed. Below the lower turbine higher Fr, respectively. For a comparison of our results with
and between the turbines no circulation was seen. Passing the data from the literature, only the work of Alves and Vascon-
lower turbine, the sparged air rose close around the impeller celos [2] studying the mixing in gas–liquid contactors agitated
shaft toward the upper turbine and entered it. Major two-phase by triple Ruston turbines in the flooding regime was available.
circulation appeared in the horizontal plane of the turbine and Comparing both results it was found that their suggested cor-
above it. With further increasing of the gas flow rate the upper relation actually represented the LFT for the lower turbine as
turbine (Du2) also became flooded. As depicted in Fig. 7, the can be seen in Fig. 8. Their results are in very good agreement
upper turbine achieved flooding transition at somewhat higher with ours. Such a basis can support the evaluation of why mod-
gas flow rates compared to single turbine stirring. On that basis, elling of mixing and dimensionless correlation of mixing time
taking the single-impeller criterion as a guide for the prediction can be possible even in a flooding regime (the lowest impeller) –
of lower impeller flooding in dual-turbine stirring seemed to be a since the middle and upper impellers are still operating under a
satisfactory estimation for practical use, which is also indicated loading regime. In [23] a qualitative observation of a two-phase
in [3,19]. flow field in a lab scale vessel with a four-stage turbine impeller
When three equal turbines are mounted on the same shaft, is presented. Only at very low impeller speeds and high gas
spaced by at least 1.5D, the lowest impeller will become flooded flow rates were all the turbines flooded; with the increasing of
long before the upper ones have ceased to function properly. the impeller speed the turbines became correspondingly loaded,
Based on visual observation it was clearly seen that the turbines which is in good agreement with our findings.
became flooded in correspondence with the increasing of the gas All experimental data of individual impeller LFT were cor-
flow rate. When the lower one was flooded the two-phase circu- related using Eq. (3) with adequate parameters shown in Table 2
lation in the liquid bulk was very similar to that in dual turbine the corresponding curves are depicted in Fig. 9. In general,
stirring. Major two-phase circulation was found in the horizon- major differences concerning gas dispersing ability can be found
tal planes of the middle and upper turbine. Further increasing of between the lower, middle and upper turbines. Comparing the
the gas flow rate led to the flooding of the middle turbine. When results of individual impeller LFTs of triple-, dual- and single-
these two turbines were flooded no circulation was seen below
the middle turbine and the liquid bulk remained undispersed. In Table 2
Corresponding coefficients to Eq. (3)
this situation the sparged air entered the upper turbine actually
undispersed. Discharging two-phase flow was still able to pre- Impeller configuration k1 k2 R
serve its own circulation. Finally, with the increasing of gas flow Single turbine (Single) 0.5007 0.7710 0.974
rates the upper turbine became flooded. In Fig. 8 the LFT for Dual: upper turbine (Du2) 0.5387 0.6133 0.989
the lower (Tr1), middle (Tr2) and upper turbine (Tr3) is shown. Dual: lower turbine (Du1) 0.3337 0.7962 0.980
At the lower turbine LFT occurred changing from VC, 1L, 2L Triple: upper turbine (Tr3) 0.5829 0.4371 0.989
to RC at Fr < 0.09 and from  3-3 to RC structure at higher Fr,
Triple: middle turbine (Tr2) 0.4901 0.6003 0.986
Triple: lower turbine (Tr1) 0.2990 0.6437 0.965
while the middle and upper impeller became flooded changing
94 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95

believe that the data could be valuable as a benchmark test for

CFD analysis that could be applicable to industrial size vessels.

4. Conclusions

The paper presents the individual impeller flooding condi-

tions in air dispersing into water stirred with a multiple-turbine
impeller. Based on local detection of the gas phase in a turbine-
discharge two-phase flow and appropriate procedure, it was
shown that the loading to flooding transition occurred via the
development of different gas filled cavity-structures irrespective
of single- or multiple-turbine stirring. This is the first time that
individual impeller flooding conditions with dual- and triple-
turbine impellers in a pilot-size aerated stirred vessel have been
detected and classified in a flow regime map.
The experimental results of flooding recognition in single-
turbine stirring were in good agreement with comparable ones
found in the literature. Comparing our experimental data with
data from the literature it can be seen that the loading to flooding
Fig. 9. Individual impeller flooding conditions; comparison of single, dual and
transition appears at somewhat higher Fl values in stirring in
triple-impeller stirring. larger vessels than ours and at lower Fl values in smaller ones.
Flooding transition of the lower turbine in dual-turbine stir-
ring occurred at lower gas flow rates than that of single-turbine
turbine stirring, the following conclusions can be drawn: the stirring, while the upper impeller remained in a loading regime
upper turbine of the triple turbines achieved the highest Fl val- in spite of marked increasing until it become flooded.
ues when it became flooded. From this point of view a triple In triple-impeller stirring the turbines became flooded in cor-
impeller as a unit is capable of dispersing the largest amount of respondence with the increasing of the gas flow rate. The lower
air, followed by dual impeller and single impeller, respectively. impeller became flooded first, followed by the flooding of the
Furthermore, the LFT of the lower turbine in dual turbines coin- middle turbine and finally the upper turbine, respectively. Con-
cides with the LFT of the lower turbine in triple turbines, as well sidering the composite impeller as a unit, the triple turbine
as the coincidence of the LFT of the middle turbine in triple tur- impeller was capable of dispersing the largest amount of air
bines with the LFT of the upper turbine in dual turbines. Our and the single impeller the smallest amount.
experimental data are consistent with the flooding relationship Experimental data on the loading to flooding transition of
suggested in [28] as well as with the expectations in [31] by the lower turbine with dual turbines coincide with those of the
which the LFT of the upper impellers would occur at a Fl num- lower turbine in triple turbines, as well as the data of the upper
ber about twice that which floods the lowest. turbine in dual-turbine stirring with those of the middle turbine
To avoid dead zones at the bottom part of the vessel it is in triple-turbine stirring. These findings can be explained by
important to know the flooding timing. By increasing the gas stable independent flow circulations of each individual impeller.
flow rate, the bottom impeller became flooded first, followed by Our results were consistent in all cases with those for which
the middle one and finally the upper. Therefore, the lowest tur- experimental data or correlations can be found in the literature.
bine was found to be the most critical for ensuring good overall
performance of a reactor which becomes flooded at remark- Acknowledgments
ably lower Fl values compared to that of single turbine stirring.
Knowing the flooding point of the upper impellers could be use- This work was financially supported by the Slovenian Min-
ful in some special occasions when the lower impeller is already istry of Higher Education, Science and Technology under con-
flooded, such as micromixing effects in an aerated tank [15], tract No. P2-162.
the effect of aeration on the mixing number [11], use of mixing
time correlation in multiple impeller stirring [2] or the use of a
References
specially composed multistage impeller working at a high gas
flooding rate [30]. [1] V. Abrardi, G. Rovero, G. Baldi, S. Sicardi, R. Conti, Hydrodynamics
At this point it should be emphasized that all experiments of a gas-liquid reactor stirred with a multiple-impeller system, Trans.
were carried out in an air/water system. The presented results Inst. Chem. Eng. A 68 (1990) 516.
could therefore serve only as an orientation for the practical [2] S.S. Alves, J.M.T. Vasconcelos, Mixing in the gas–liquid contactors
analysis of fermenters at the beginning of the process when the agitated by multiple turbines in the flooding regime, Chem. Eng. Sci.
50 (14) (1995) 2355.
fermentation broth behaves in an aqueous manner. Moreover, the [3] A. Bakker, J.M. Smith, K.J. Myers, How to disperse gases in liquids,
pilot-size vessel had a fixed D/T ratio which can not be taken a Reprint from Chemical Engineering, December 1994, by McGraw-Hill
priori when subjected to scaled up design. However, the authors Inc.
 A. Bombač, I. Žun / Chemical Engineering Journal 116 (2006) 85–95 95

[4] A. Bombač, Effects of geometrical parameters on Newton number in an [22] M. Nocentini, F. Magelli, G. Pasquali, Some comments on the flooding-
aerated stirred tank, J. Mech. Eng. 44 (3–4) (1998) 105. loading transition for gas–liquid vessels stirred with Rushton turbines,
[5] A. Bombač, I. Žun, M. Žumer, J. Turk, Zbornik Kuhljevi dnevi 1996, Chem. Eng. Res. Des. 66 (1988) 378.
Slovensko društvo za mehaniko, Ljubljana, 1996, pp. 193–198 (in [23] M. Nocentini, F. Magelli, G. Pasquali, D. Fajner, A fluid dynamic study
Slovene). of a gas–liduid, non-standard vessel stirred by multiple impeller, Chem.
[6] A. Bombač, I. Žun, B. Filipič, M. Žumer, Gas-filled cavity structures Eng. J. 37 (1988) 53.
and local void fraction distribution in aerated stirred vessel, AIChE J. [24] A. Paglianti, S. Pintus, M. Giona, Time-series analysis approach for
43 (11) (1997) 2921. the identification of flooding/loading transition in gas–liquid stirred tank
[7] A. Bombač, I. Žun, Gas-filled cavity structures and local void fraction reactors, Chem. Eng. Sci. 55 (2000) 5793.
distribution in vessel with dual impellers, Chem. Eng. Sci. 55 (2000) [25] A. Paglianti, Simple model to evaluate loading/flooding transition in
2995. aerated vessels stirred by Rushton disc turbines, Can. J. Chem. Eng. 80
[8] A. Bombač, I. Žun, A simple method for detecting individual impeller (2002) 660.
flooding of dual Rushton impellers, in: Proceedings of the European [26] E.J. Paul, V.A. Atiemo-Obeng, S.M. Kresta, Handbook of Industrial
Conference on Mixing X, Elsevier Science B.V., 2000, pp. 469–476. Mixing, John Wiley and Sons, Hoboken, NJ, 2004.
[9] A. Bombač, I. Žun, Flooding recognition methods in a turbine stirred [27] M. Roustan, M.Z.A. Bruxelmane, CHISA, Prague, (1981), paper B3.3.
vessel, J. Mech. Eng. 48 (12) (2002) 663. [28] M. Roustan, Power consumed by the Rushton turbines in non standard
[10] P.R. Gogate, A.A.C.M. Beenackers, A.B. Pandit, Multiple-impeller sys- vessel under gassed conditions, in: Proceedings of the 5th European
tems with a special emphasis on bioreactors: a critical review, Biochem. Conference on Mixing, BHRA, The Fluid Engineering Centre Cranfield,
Eng. J. 6 (2) (2000) 109. Bedford, UK, 1985, pp. 127–141.
[11] F. Guillard, C. Trägårdh, Mixing in industrial Rushton turbine-agitated [29] J.H. Rushton, J.J. Bimbinet, Hold-up and flooding in air–water mixing,
reactors under aerated conditions, Chem. Eng. Process 42 (5) (2003) Can. J. Chem. Eng. 46 (1968) 16.
373. [30] R. Sardeing, F. Ferrand, M. Poux, P. Avrillier, C. Xuereb, Hydro-
[12] H.J. Henzler, Verfarentechnishe Auslegungsunterlagen fur Ruhrbehalter dynamics and gas dispersion characterization in a system equipped
als Fermenter, Chem. Ing. Tech. 52 (5) (1982) 461. with a new gas-inducing impeller, Chem. Eng. Technol. 26 (1) (2003)
[13] V. Hudcova, A.W. Nienow, W. Haozhung, L. Huoxing, On the effect of 31.
liquid height on the flooding/loading transition, Chem. Eng. Sci. 42 (2) [31] J.M. Smith, M.M.C.G. Warmoeskerken, E. Zeef, Flow conditions in
(1987) 375. vessel dispersing gases in liquids with multiple impellers, in: C.S.
[14] V. Hudcova, V. Machon, A.W. Nienow, Gas-liquid dispersion with dual Ho, J.Y. Oldshue (Eds.), Biotechnology Processes, AIChE, 1987,
Rushton turbine impellers, Biotechnol. Bioeng. 34 (1989) 617. p. 107.
[15] W.W. Lin, D.J. Lee, Micromixing effects in aerated stirred tank, Chem. [32] G.B. Tatterson, G.L. Morison, Effect to tank to impeller diameter ratio
Eng. Sci. 52 (21–22) (1997) 3837. on flooding transition for disk turbines, AIChE J. 33 (10) (1987) 1751.
[16] W. Lu, H. Chen, Flooding and critical impeller speed for gas dispersion [33] M.M.C.G. Warmoeskerken, J.M. Smith, The flooding transition with
in aerated turbine-agitated vessels, Chem. Eng. J. 33 (1986) 57. gassed Rushton turbines, Fluid Mixing II, Symposium Series No. 89,
[17] Wei-Ming Lu, Shin-Jon Ju, Cavity configuration, flooding and pumping 1985, 59.
capacity of disc-type turbines in aerated stirred vessel, Chem. Eng. Sci. [34] M.M.C.G. Warmoeskerken, J.M. Smith, Flooding of disk turbines in
44 (2) (1989) 333. gas–liquid dispersions: a new description of the phenomenon, Chem.
[18] R. Mezaki, M. Mochizuki, K. Ogawa, Engineering Data on Mixing, Eng. Sci. 40 (11) (1985) 2063.
Elsevier, Amsterdam, 2000. [35] J.A. Wiedmann, Zum Ueberflutunverhalten Zwei- und Dreiphasig
[19] N. Harnby, M.F. Edwards, A.W. Nienow, Mixing in the Process Indus- Betriebener Rhuerreaktoren, Chem. Ing. Tech. 55 (9) (1983) 689.
tries, 2nd ed., Butterworths Heinemann, Oxford, 1992. [36] M. Zlokarnik, Ruhrleistung in begasten flussigkeiten, Chem. Ing. Tech.
[20] E. Mikulcova, V. Kudrna, J. Vlcek, Bestimmung der Grenzbedinungen 45 (10a) (1973) 689.
fuer die Ueberflutung des Ruehrers K1, Scientific paper of the Institute [37] I. Žun, A. Bombač, An application of local void fraction measurements
of Chemical Technology, Prague, 1967, p. 167. in discharge flow of single and dual Rushton impellers, in: Proceedings
[21] A.W. Nienow, M.M.C.G. Warmoeskerken, J.M. Smith, M. Konno, On of the Mixing IX – Multiphase Systems, GFGP, Paris-Marne La Vallee,
the flooding/loading transition and the complete dispersal condition in 1997, pp. 153–160.
aerated vessels agitated by a Rushton-turbine, in: Proceedings of the 5th [38] I. Žun, B. Filipič, M. Perpar, A. Bombač, Phase discrimination and void
European Conference on Mixing, BHRA, The Fluid Engineering Centre fraction measurements via genetic algorithms, Rev. Sci. Instrum. 66 (10)
Cranfield, Bedford, UK, 1985, pp. 143–154. (1995) 5055.